the tidyverse - genome analysis wiki · data tables the "data.table" package exist to...

dplyr(and the tidyverse)

Matthew Flickinger, Ph.D.

CSG Tech Talk

University of Michigan

July 12, 2017

The Tidyverse

http://varianceexplained.org/r/teach-tidyverse/

Tidyverse

Very popular, widely used

Prioritize data analysis rather

than computer science

Enable learners to become

productive more quickly

Encourages readable code

Dplyr motivation

Analysists spend a lot of time manipulating and

summarizing data

Base R provides many functions for this, but

the syntax is sometimes verbose or "ugly"

the functions can be slow for big data

dplyr exists to make code easier to read and faster

Install and load dplyr

Install via tidyverse

install.packages("tidyverse")

library(tidyverse)

OR install directly

install.packages("dplyr")

library(dplyr)

This guide assumes you're running dplyr 0.7.1 (released

June 22, 2017)

Sample data

Examples use a data set containing all out-bound flights

from NYC in 2013

Available as an R package

install.packages("nycflights13")

library(nycflights13)

"flights" table

> flightsSource: local data frame [336,776 x 19]

year month day dep_time sched_dep_time dep_delay arr_time sched_arr_time(int) (int) (int) (int) (int) (dbl) (int) (int)

1 2013 1 1 517 515 2 830 8192 2013 1 1 533 529 4 850 8303 2013 1 1 542 540 2 923 8504 2013 1 1 544 545 -1 1004 10225 2013 1 1 554 600 -6 812 8376 2013 1 1 554 558 -4 740 7287 2013 1 1 555 600 -5 913 8548 2013 1 1 557 600 -3 709 7239 2013 1 1 557 600 -3 838 84610 2013 1 1 558 600 -2 753 745.. ... ... ... ... ... ... ... ...Variables not shown: arr_delay (dbl), carrier (chr), flight (int), tailnum

(chr), origin (chr), dest (chr), air_time (dbl), distance (dbl), hour (dbl),minute (dbl), time_hour (time)

Basic single-table dplyr verbs

Basic dplyr verbs

filter() – keep rows matching desired properties

select() – choose which columns you want to extract

arrange() – sort rows

mutate() – create new columns

summarize() – collapse rows into summaries

group_by() – operate on subsets of rows at a time

dplyr verb properties

Always take a data source as the first parameter

Returns a new data object, never updates/replace original

Specify columns as unquoted strings (symbols)

Filtering Rows

Find all flights to Detroit (DTW) in June (2013)

dplyr

filter(flights, dest=="DTW" & month==6)

Base R

flights[flights$dest=="DTW" & flights$month==6, ]

subset(flights, dest=="DTW" & month==6)

Selecting columns

Select specific columns

select(flights, dep_time, arr_time, carrier)

Exclude columns

select(flights, -year, -tailnum)

Select column range

select(flights, month:dep_delay)

Selecting columns … part 2

select(flights, starts_with("d"))

select(flights, ends_with("time"))

select(flights, contains("arr"))

select(flights, -starts_with("d"))

select(flights, flight, everything())

See "?select" for complete list and examples

Verb composition via magrittr

What if we want to both filter and select?

We could create variables for each intermediate step

filtered <- filter(flights, dest=="DTW")

select(filtered, carrier)

Or we could nest the calls

select(filter(flights, dest=="DTW"), carrier)

But these can get messy

Verb composition via magrittr

The magrittr package introduces the "%>%" operator to "pipe" data into function

Similar to the unix pipe operator

grep apple fruit.txt | head -50 | cut –f3 | sort | uniq –c

The %>% operator passes the result from the left side to the first argument of the right side

a(b(c(x)) x %>% c() %>% b() %>% a()

flights %>% filter(dest=="DTW") %>% select(carrier)

Sort data

Use arrange() to sort your rows

flights %>% arrange(sched_dep_time)

Use desc() to reverse the sort order of a column

flights %>% arrange(month, desc(day))

You can sort on functions of variables

flights %>% arrange(desc(dep_time-sched_dep_time ))

Create new variables

Mutate allows you to create columns using existing values

flights %>% mutate(speed = distance/(air_time/60)) %>% arrange(desc(speed)) %>% select(flight, speed)

Remember, changes are not saved to "flights", be sure to

save the result if you want to use it later

new_flights <- flights %>% mutate(...)

Use new variables right away

The parameters to mutate are processed in the order they

appear

flights %>% mutate(dist_km = distance * 1.61, hours = air_time / 60, kph = dist_km/hours ) %>%

select(flight, kph)

Be careful! You can overwrite existing variables

Summarize data

You generally use summarize() to reduce the number of

rows in your data by specifying summary functions for

each of the columns

flights %>% filter(!is.na(arr_delay)) %>% summarize(avg_arr_delay = mean(arr_delay))

Useful summary functions:

mean(), median(), var(), sd(), min(), max(), first(), last(),

n(), n_distinct()

Grouping data

Often you want to perform summaries for groups of rows

at a time

The group_by() function allows your to specify columns

that define groups

Functions like mutate() and summarize() are

performed for each group

group_by() + summarize() example

flights %>% filter(!is.na(arr_delay)) %>% group_by(carrier) %>% summarize(avg_arr_delay = mean(arr_delay))

# A tibble: 16 x 2carrier avg_arr_delay

<chr> <dbl>1 9E 7.37966922 AA 0.36429093 AS -9.93088864 B6 9.4579733...

Combining group_by() with

transformations

mutate()

Will not change the number of rows

Functions like max() will return the max for each group

summarize()

Returns one row per group

You must tell summarize how to collapse all non-grouped

columns that you want

group_by() + mutate() example

flights %>% filter(!is.na(arr_delay)) %>% group_by(carrier) %>% mutate(avg_arr_delay = mean(arr_delay)) %>% select(carrier, arr_delay, avg_arr_delay)

# A tibble: 327,346 x 3# Groups: carrier [16]

carrier arr_delay avg_arr_delay<chr> <dbl> <dbl>

1 UA 11 3.55801112 UA 20 3.55801113 AA 33 0.36429094 B6 -18 9.4579733...

Merging data

What are these carrier codes?

flights %>% filter(!is.na(arr_delay)) %>% group_by(carrier) %>% summarize(avg_arr_delay = mean(arr_delay))

# A tibble: 16 x 2carrier avg_arr_delay

<chr> <dbl>1 9E 7.37966922 AA 0.36429093 AS -9.93088864 B6 9.4579733...

"airlines" table

> airlines

# A tibble: 16 x 2

carrier name

<chr> <chr>

1 9E Endeavor Air Inc.

2 AA American Airlines Inc.

3 AS Alaska Airlines Inc.

4 B6 JetBlue Airways

5 DL Delta Air Lines Inc.

...

Join flights to airlines

> flights %>% filter(!is.na(arr_delay)) %>% group_by(carrier) %>% summarize(avg_arr_delay = mean(arr_delay)) %>%left_join(airlines)

Joining, by = "carrier"

# A tibble: 16 x 3

carrier avg_arr_delay name

<chr> <dbl> <chr>

1 9E 7.3796692 Endeavor Air Inc.

2 AA 0.3642909 American Airlines Inc.

3 AS -9.9308886 Alaska Airlines Inc.

4 B6 9.4579733 JetBlue Airways

Types of joins (merges)

Types of joins (merges)

Types of joins (merges)

Types of joins (merges)

Non-merging joins

These "joins" do not add any new columns to your data

They useful for subsetting with multi-column matches

semi_join()

Only keep rows in left table with matches in right

anti_join()

Drop rows in left table with matches in right

Join on

By default the join commands will join two tables based

on all matching column names

You can control the joining by specifying the column

names

flights %>% inner_join(planes) %>% nrow

flights %>% inner_join(planes, "tailnum") %>% nrow

Other dplyr functions

Subsetting observations

distinct()

Return unique rows

flights %>% distinct(tailnum, carrier)

count()

count rows with unique values of selected columns

flights %>% count(carrier)

sample_n()

Randomly choose n rows

flights %>% sample_n(3)

Transformation functions

summarize_all()/mutate_all()

Apply function to all non-grouped columns

summarize_at()/mutate_at()

Apply function to chosen columns

flights %>% summarize_at(vars(ends_with("time")), mean, na.rm=T)

summarize_if()/mutate_if()

Apply function to matching columns

flights %>% summarize_if(is.numeric, funs(mean, var), na.rm=T)

See vars() and funs() for passing multiple columns or functions

Other functions

lead()/lag()

Get value just after/before current value

x<-1:5; x; lead(x); lag(x)

coalesce()

Returns first non-NA value from vectors

coalesce(c(NA,2,NA), c(1, NA, NA), 3)

n()/n_distinct()

Number of (distinct) values is a vector

Can only be used within summarize(), mutate(), filter()

flights %>% group_by(tailnum) %>% summarize(flights=n(), routes=n_distinct(flight))

[1] 1 2 3 4 5[1] 2 3 4 5 NA[1] NA 1 2 3 4

[1] 1 2 3

Other functions

recode()

Replace values in vector with different values

recode(letters[1:5], b="boo")

case_when()

Alternative to nested ifelse() calls

x <- 1:50

case_when(

x %% 35 == 0 ~ "fizz buzz",

x %% 5 == 0 ~ "fizz",

x %% 7 == 0 ~ "buzz",

TRUE ~ as.character(x)

)

[1] "a" "boo" "c" "d" "e"

[1] "1" "2" "3" "4" [5] "fizz" "6" "buzz" "8" [9] "9" "fizz" "11" "12" [13] "13" "buzz" "fizz" "16"

Combining data frames

bind_rows()

Stack two data frames on top of each other (should have the

same number of columns)

bind_columns()

Place two data frames next to each other (should have the

same number of rows) – no merge-able columns

intersect(), union(), setdiff()

For rows shared or exclusive to data frames

Other benefits of dplyr

Generic data interface

In all these examples we've been working with basic R

data.frames

However, dplyr can act as a front end to other data types

This include databases (via SQL) and data.tables

"Tibbles"

> class(flights)

[1] "tbl_df" "tbl" "data.frame"

dplyr works with objects of class "tbl" (pronounced tibble)

Here we have a wrapped data.frame

Data tables

The "data.table" package exist to make data frame like

structures that are faster and more efficient to work with

The "data.table" package overload the subset operator "["

to allow for grouping and subsetting in a non-standard way

If you load the "dtplyr" package, you can use the nicer

dplyr functions to work with data tables as well

(Prior to dplyr 0.5.0, the data tables functions were in the

same package and didn't need to be loaded separately)

Programming with dplyr

Writing functions with dplyr

# WORKS

flights %>%

group_by(carrier) %>%

summarize(delay=mean(arr_delay, na.rm=T))

# DOESN’T WORK

f <- function(x) {

flights %>%

group_by(x) %>%

summarize(delay=mean(arr_delay, na.rm=T))

}

f(carrier) # ERROR: Column `x` is unknown

Quosures

f <- function(x) {

flights %>% group_by(!!x) %>%

summarize(delay = mean(arr_delay, na.rm=T))

}

f(quo(carrier))

g <- function(x) {

x <- enquo(x)

flights %>% group_by(!!x) %>%

summarize(delay = mean(arr_delay, na.rm=T))

}

g(carrier)

# A tibble: 16 x 2carrier delay

<chr> <dbl>1 9E 7.37966922 AA 0.3642909...

Rename output (:=)

h <- function(x) {

x <- enquo(x)

outname <- paste(quo_name(x), "delay", sep="_")

flights %>% group_by(!!x) %>%

summarize(!!outname := mean(arr_delay, na.rm=T))

}

h(carrier) # A tibble: 16 x 2carrier carrier_delay<chr> <dbl>

1 9E 7.37966922 AA 0.3642909...

"Data Wrangling Cheat Sheet"

you %>%select(interesting_dataset) %>%summarize(features) %>%test(hypothesis) %>%profit() %>%the_end()

nycflights13 tables